Tuesday, February 16, 2016

Trump to Political Science: Drop Dead, Part I

OK, enough about the Supreme Court, the death of a historically important Justice, and the future of advice and consent.  Scalia’s death, like everything else, is really about Donald Trump.  As Roll Call’s Melinda Hennenberger suggests, maybe this will be the thing that finally does in Trump’s campaign.  This would be more compelling if we hadn’t heard the same thing a dozen times by now.

And it isn’t just pundits.  Worse, perhaps, is the denial among political scientists.  Political science is in a crisis now, brought on by The Donald.  Trump is winning.  He still might lose.  In fact, PredictWise currently puts his chances just under 50%, but if you had asked any political scientist about Trump’s chances last summer, answers would have ranged from zero to “stop wasting my time with stupid questions.”  With a New Hampshire victory under his belt, a still solid polling lead, leads heading into the South Carolina and Nevada contests, and a demoralized Republican “establishment,” try asking a political scientist right now how sure they are that Trump will lose.  If they say they are as sure as they were last summer, ask them how much money they have riding on it.  You can buy “shares” of Trump losing for around $.55 on PredictIt, and that pays out at $1 when it happens.  Of course, if Trump really does lose, it will have to be soon, and you’ll be able to sell at around $.95 within a couple of months.  That’s nearly doubling your money in a matter of months.  Anyone not buying those shares is either irrational or less sure of Trump’s loss than we all were last summer.  Let’s face it.  Political science underestimated Trump.  And most still don’t want to admit how close he is to pants-ing the discipline of political science.

Political science, then, is in a state of denial about Trump.  Let’s call it “Trump denialism.”  Trump denialism takes two forms.  Type A denialism is the continued refusal to acknowledge that Trump actually has a chance of winning.  In its weaker form, Type A denialism is the slowness and grudgingness with which many observers have come to accept that he might win.  We all denied that last summer.  Anyone still denying it has a problem.  Type B denialism is the refusal to acknowledge that Trump’s unexpected strength as a candidate indicates a problem in conventional political science understandings of nomination politics.

We’ll start with Type A Trump denialism.  Let’s just be blunt here.  Trump could still lose, but he is in a pretty strong position, and none of us saw it coming.  Here are the basic facts.  Trump has maintained a persistent lead in the polls since entering the race.  There was a very brief period when he was essentially tied with Ben Carson, but that didn’t last.  For all practical purposes, Trump has led in the polls since he entered.  He just won New Hampshire.  That’s not a lock on the nomination.  Past winners of the New Hampshire primary include Hillary in 2008, John McCain in 2000, Pat Buchanan in 1996, Paul Tsongas in 1992…  You get the point.  But, that list also includes Mitt Romney in 2012, John McCain in 2008, John Kerry in 2004, Al Gore in 2000…  Winning New Hampshire doesn’t seal the deal, but combined with long-running dominance in the polls, it means you can’t write Trump off anymore.

What else should we look at?  Money?  Jeb Bush has been flooded with campaign cash and it hasn’t done him any good, but Trump has essentially unlimited money if he ever needed it.  Endorsements?  Rubio is ahead in endorsements, but very few Senators, Representatives or Governors on the Republican side have endorsed anyone.  Compare Rubio’s lead to Clinton’s on the Democratic side.  So, it is hard to see that working against him.  Party Decides apparatchiks keep expecting a flood of endorsements to an establishment-friendly candidate, and it keeps not happening.

So, Trump is way ahead in the polls, won New Hampshire, is leading in South Carolina, has unlimited money, and nobody has a significant endorsement lead on him.  Trump could still lose, but anyone saying right now that he has no chance is just in denial.  In fact, the prediction markets look like they might be underestimating his chances.  PredictWise currently gives him a 48% chance of winning.  But, they also give him an 86% chance of winning South Carolina and a 69% chance of winning Nevada.  And those aren’t independent events.  Given that, why do people put his odds below 50%?  It is hard to see a logical basis.  Essentially, it boils down to, “really?!  Trump?!”  In other words, a gut-level assessment that all of our empirical observations misled us.  Who are you going to believe?  Your gut, or those lying data?

Trump is winning, and nobody wants to admit it.  Least of all, political scientists.  He could still lose, but even if he does, we clearly underestimated him, and those still dismissing his chances are guilty of Type A Trump denialism.

What, then, can explain the persistence of Type A Trump denialism?  First, there is simple ego defense.  We all wrote Trump off as an impossibility.  Most of us thought he wouldn’t even formally declare.  We get locked into our certainty.  The longer we postpone dealing with Trump beating expectations, the longer we postpone critical self-analysis.  Just like the longer we postpone going to the dentist for that toothache, the longer we postpone that terrifying drill.  But, just like avoiding the dentist, the longer we wait, the worse it gets when we face facts.

Of course, that brings us to the second motivation behind Type A denialism.  He could still lose!  In fact, if the betting odds are correct now, he is still more likely to lose than win!  Thus, if we remain in denial, there is a high likelihood that we never have to face the fact that we underestimated him.  Instead, we get to say, “I knew it all along!”  And that’s the rub.  Trump denialism is a manifestation of wishful thinking among people who hope they will eventually get to brag about never losing faith.

Faith, though.  As in, not science.  Here is where everyone needs to read Philip Tetlock’s Expert Political Judgment.  The key issue is how we respond when a prediction comes true, and how we respond when it doesn’t come true.  When we make a wrong prediction, we look for ways to say, “I was almost right,” and that except for some small, unforeseeable factor that makes it a weird case, I would have been right.  Obviously, I’ll stick with my method and be right next time.  When our predictions turn out right, though, we never perform this kind of self-diagnosis.  Were we almost wrong?  Was there some flaw in our reasoning, but we got lucky?  Nobody wants to face that, and it is the key to understanding Type A Trump denialism.  Denialists are hoping that The Donald loses so that they can avoid ever having to perform the rigorous self-analysis showing all of the flaws in their reasoning that made them almost wrong.  They can go back to saying Trump never had a chance.  Let’s just call bullshit on that right now.

We, as a discipline, put Trump’s odds at zero.  That assessment was wrong.  Even if he loses, we clearly underestimated him.  And we need to come to grips with that, and what it means for our understanding of the electoral process.

What went wrong?  Overconfidence in crappy models, and that’s where Type B Trump denialism comes in.  There are plenty of models that would predict a Trump loss.  The 800-pound gorilla in the room is The Party Decides, which basically says that party muckety-mucks control the process by endorsing their chosen candidates, and voters do as they’re told.  Party Decides apparatchiks had obvious reasons to discount Trump.  Then, there’s the “discovery, scrutiny and decline” model from The Gamble.  Basically, that model said that a candidate like Trump might get a bubble, but it would inevitably deflate when media attention turns to criticism.  Given what happened to candidates like Herman Cain in 2012, that made sense.  Not so, this time.  There’s the basic spatial model, suggesting that primary voters are more ideological than the general electorate, and Trump’s lack of ideological coherence should hurt him with the primary.  Yeah, that didn’t happen.  The new variant on this, associated with Grossman & Hopkins, is that Republicans in particular are prone to a desire for ideological purity.  Trump wouldn’t have a chance, then.  Self-funded candidates tend to lose, and Trump promised to self-fund.  He hasn’t even needed to spend money!  That’s at least five prominent models, all suggesting that Trump was toast.  Maybe there’s something wrong with these models.  Denying that is Type B Trump denialism.


Coming soon, then, is a series of posts on how and why each of our models have failed us.  Do I have a better one?  Nope.  Not yet.  Maybe one will come out of this.  For now, though, Trump told Political Science to drop dead.  Let’s not oblige.  That means coming to grips with our errors, and looking for a way to not make them in the future.

No comments:

Post a Comment